[Bioc-devel] SummarizedExperiment: structure loss, when mixing matrix and data.frame data

Confirmed with the following sessionInfo(), satisfying biocValid()==TRUE

sessionInfo()

R Under development (unstable) (2017-11-22 r73776)

Platform: x86_64-pc-linux-gnu (64-bit)

Running under: Linux Mint 18.1


Matrix products: default

BLAS: /home/stvjc/R-35-dist/lib/R/lib/libRblas.so

LAPACK: /home/stvjc/R-35-dist/lib/R/lib/libRlapack.so


locale:

  [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C

  [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8

  [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8

  [7] LC_PAPER=en_US.UTF-8       LC_NAME=C

  [9] LC_ADDRESS=C               LC_TELEPHONE=C

[11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C


attached base packages:

[1] parallel  stats4    stats     graphics  grDevices utils     datasets

[8] methods   base


other attached packages:

[1] SummarizedExperiment_1.9.2 DelayedArray_0.5.5

[3] matrixStats_0.52.2         Biobase_2.39.0

[5] GenomicRanges_1.31.1       GenomeInfoDb_1.15.1

[7] IRanges_2.13.4             S4Vectors_0.17.10

[9] BiocGenerics_0.25.0


loaded via a namespace (and not attached):

  [1] lattice_0.20-35         bitops_1.0-6            grid_3.5.0

  [4] zlibbioc_1.25.0         XVector_0.19.1          Matrix_1.2-12

  [7] tools_3.5.0             RCurl_1.95-4.8          compiler_3.5.0

[10] GenomeInfoDbData_0.99.2

On Sun, Nov 26, 2017 at 7:09 AM, Felix Ernst <felix.ernst at ulb.ac.be> wrote:

Hi all,

I got different results constructing a SummarizedExperiment in 3.6 and
3.7. My question is, whether this is intentional or a bug.

library(GenomicRanges)
library(SummarizedExperiment)

nrows <- 200; ncols <- 6
counts <- matrix(runif(nrows * ncols, 1, 1e4), nrows)
colnames(counts) <- LETTERS[1:6]
rownames(counts) <- 1:nrows
counts2 <- counts-floor(counts)
rowRanges <- GRanges(rep(c("chr1", "chr2"), c(50, 150)),
                      IRanges(floor(runif(200, 1e5, 1e6)), width=100),
                      strand=sample(c("+", "-"), 200, TRUE),
                      feature_id=sprintf("ID%03d", 1:200))
colData <- DataFrame(Treatment=rep(c("ChIP", "Input"), 3),
                      row.names=LETTERS[1:6])

se <- SummarizedExperiment(assays=list(counts=counts),
                            rowRanges=rowRanges,
                            colData=colData)

str(assays(se)$counts)
assays(se)$counts2 <- as.data.frame(counts2)
str(assays(se)$counts)

On a Windows 10 R3.4.2 Bioc 3.6 this produces:
num [1:200, 1:6] 8815 6314 1945 6185 5935 ...
  - attr(*, "dimnames")=List of 2
   ..$ : chr [1:200] "1" "2" "3" "4" ...
   ..$ : chr [1:6] "A" "B" "C" "D" ...
  num [1:200, 1:6] 8815 6314 1945 6185 5935 ...
  - attr(*, "dimnames")=List of 2
   ..$ : chr [1:200] "1" "2" "3" "4" ...
   ..$ : chr [1:6] "A" "B" "C" "D" ...

On Ubuntu 17.10 R-devel r73779 Bioc3.7  this produces
num [1:200, 1:6] 8636 7040 9275 4821 2475 ...
  - attr(*, "dimnames")=List of 2
   ..$ : chr [1:200] "1" "2" "3" "4" ...
   ..$ : chr [1:6] "A" "B" "C" "D" ...
  num [1:1200] 8636 7040 9275 4821 2475 ...

Somehow the structure is lost.

This happens, if I mix matrix and data.frame data, and doesn?t, if I use
only matrices. The man page defines matrix-like objects,
which a data.frame is (isn?t it?) and the behavior is different from
Bioc3.6 to Bioc3.7.

I can rule out that this is a Windows/Linux thing, because the Travis
build error, which pointed to an difference in the first place,
didn?t occur with bioc-release, just with bioc-devel.

Thanks for any advice and suggestions.

Felix

         [[alternative HTML version deleted]]

	[[alternative HTML version deleted]]